Avoiding loops while accessing pixel information of an image - java

I'm trying to copy the RGB values of an image to a 2D array. At the moment I'm using two nested for loops. My question is, is it possible to use arraycopy or any other method to make the operation faster?
The below code works fine, I just would like to optimize it.
private LSHColor[][] pixels = new LSHColor[w][h];
public LSHImage(BufferedImage img) {
pixels = new LSHColor[img.getWidth()][img.getHeight()];
for (int y = 0; y < img.getHeight(); y++)
for (int x = 0; x < img.getWidth(); x++)
pixels[x][y] = LSHColor.fromRGB(img.getRGB(x, y));
}

arraycopy is not going to help for two reasons. First reason is that you are not making a copy: you are creating an array based on the content of another array. On top of that, your pixels[][] is in column-major order, meaning that pixels of the same column are next to each other in memory, while images are stored in row-major order.
One thing that you can optimize is getting your pixels from the image. Rather than reading them one at a time, you can get them all at once, and then use your loops for processing, like this:
public LSHImage(BufferedImage img) {
int w = img.getWidth();
int h = img.getHeight();
pixels = new LSHColor[w][h];
int[] raw = img.getRGB(0, 0, w, h, null, 0, w);
for (int y = 0; y < img.getHeight(); y++)
for (int x = 0; x < img.getWidth(); x++)
pixels[x][y] = LSHColor.fromRGB(raw[w*y+x]);
}
This makes a single call to getRGB instead of making Width*Height calls. Tradeoff here is that you use some temporary memory for the image, which may be prohibitive for larger images.

Related

Java Convolution

Hi I am in need of some help. I need to write a convolution method from scratch that takes in the following inputs: int[][] and BufferedImage inputImage. I can assume that the kernel has size 3x3.
My approach is to do the follow:
convolve inner pixels
convolve corner pixels
convolve outer pixels
In the program that I will post below I believe I convolve the inner pixels but I am a bit lost at how to convolve the corner and outer pixels. I am aware that corner pixels are at (0,0), (width-1,0), (0, height-1) and (width-1,height-1). I think I know to how approach the problem but not sure how to execute that in writing though. Please to aware that I am very new to programming :/ Any assistance will be very helpful to me.
import java.awt.*;
import java.awt.image.BufferedImage;
import com.programwithjava.basic.DrawingKit;
import java.util.Scanner;
public class Problem28 {
// maximum value of a sample
private static final int MAX_VALUE = 255;
//minimum value of a sample
private static final int MIN_VALUE = 0;
public BufferedImage convolve(int[][] kernel, BufferedImage inputImage) {
}
public BufferedImage convolveInner(double center, BufferedImage inputImage) {
int width = inputImage.getWidth();
int height = inputImage.getHeight();
BufferedImage inputImage1 = new BufferedImage(width, height, BufferedImage.TYPE_INT_ARGB);
//inner pixels
for (int x = 1; x < width - 1; x++) {
for (int y = 1; y < height - 1; y ++) {
//get pixels at x, y
int colorValue = inputImage.getRGB(x, y);
Color pixelColor = new Color(colorValue);
int red = pixelColor.getRed() ;
int green = pixelColor.getGreen() ;
int blue = pixelColor.getBlue();
int innerred = (int) center*red;
int innergreen = (int) center*green;
int innerblue = (int) center*blue;
Color newPixelColor = new Color(innerred, innergreen, innerblue);
int newRgbvalue = newPixelColor.getRGB();
inputImage1.setRGB(x, y, newRgbvalue);
}
}
return inputImage1;
}
public BufferedImage convolveEdge(double edge, BufferedImage inputImage) {
int width = inputImage.getWidth();
int height = inputImage.getHeight();
BufferedImage inputImage2 = new BufferedImage(width, height, BufferedImage.TYPE_INT_ARGB);
//inner pixels
for (int x = 0; x < width - 1; x++) {
for (int y = 0; y < height - 1; y ++) {
//get pixels at x, y
int colorValue = inputImage.getRGB(x, y);
Color pixelColor = new Color(colorValue);
int red = pixelColor.getRed() ;
int green = pixelColor.getGreen() ;
int blue = pixelColor.getBlue();
int innerred = (int) edge*red;
int innergreen = (int) edge*green;
int innerblue = (int) edge*blue;
Color newPixelColor = new Color(innerred, innergreen, innerblue);
int newRgbvalue = newPixelColor.getRGB();
inputImage2.setRGB(x, y, newRgbvalue);
}
}
return inputImage2;
}
public BufferedImage convolveCorner(double corner, BufferedImage inputImage) {
int width = inputImage.getWidth();
int height = inputImage.getHeight();
BufferedImage inputImage3 = new BufferedImage(width, height, BufferedImage.TYPE_INT_ARGB);
//inner pixels
for (int x = 0; x < width - 1; x++) {
for (int y = 0; y < height - 1; y ++) {
//get pixels at x, y
int colorValue = inputImage.getRGB(x, y);
Color pixelColor = new Color(colorValue);
int red = pixelColor.getRed() ;
int green = pixelColor.getGreen() ;
int blue = pixelColor.getBlue();
int innerred = (int) corner*red;
int innergreen = (int) corner*green;
int innerblue = (int) corner*blue;
Color newPixelColor = new Color(innerred, innergreen, innerblue);
int newRgbvalue = newPixelColor.getRGB();
inputImage3.setRGB(x, y, newRgbvalue);
}
}
return inputImage3;
}
public static void main(String[] args) {
DrawingKit dk = new DrawingKit("Compositor", 1000, 1000);
BufferedImage p1 = dk.loadPicture("image/pattern1.jpg");
Problem28 c = new Problem28();
BufferedImage p5 = c.convolve();
dk.drawPicture(p5, 0, 100);
}
}
I changed the code a bit but the output comes out as black. What did I do wrong:
import java.awt.*;
import java.awt.image.BufferedImage;
import com.programwithjava.basic.DrawingKit;
import java.util.Scanner;
public class Problem28 {
// maximum value of a sample
private static final int MAX_VALUE = 255;
//minimum value of a sample
private static final int MIN_VALUE = 0;
public BufferedImage convolve(int[][] kernel, BufferedImage inputImage) {
int width = inputImage.getWidth();
int height = inputImage.getHeight();
BufferedImage inputImage1 = new BufferedImage(width, height, BufferedImage.TYPE_INT_ARGB);
//for every pixel
for (int x = 0; x < width; x ++) {
for (int y = 0; y < height; y ++) {
int colorValue = inputImage.getRGB(x,y);
Color pixelColor = new Color(colorValue);
int red = pixelColor.getRed();
int green = pixelColor.getGreen();
int blue = pixelColor.getBlue();
double gray = 0;
//multiply every value of kernel with corresponding image pixel
for (int i = 0; i < 3; i ++) {
for (int j = 0; j < 3; j ++) {
int imageX = (x - 3/2 + i + width) % width;
int imageY = (x -3/2 + j + height) % height;
int RGB = inputImage.getRGB(imageX, imageY);
int GRAY = (RGB) & 0xff;
gray += (GRAY*kernel[i][j]);
}
}
int out;
out = (int) Math.min(Math.max(gray * 1, 0), 255);
inputImage1.setRGB(x, y, new Color(out,out,out).getRGB());
}
}
return inputImage1;
}
public static void main(String[] args) {
int[][] newArray = {{1/9, 1/9, 1/9}, {1/9, 1/9, 1/9}, {1/9, 1/9, 1/9}};
DrawingKit dk = new DrawingKit("Problem28", 1000, 1000);
BufferedImage p1 = dk.loadPicture("image/pattern1.jpg");
Problem28 c = new Problem28();
BufferedImage p2 = c.convolve(newArray, p1);
dk.drawPicture(p2, 0, 100);
}
}
Welcome ewuzz! I wrote a convolution using CUDA about a week ago, and the majority of my experience is with Java, so I feel qualified to provide advice for this problem.
Rather than writing all of the code for you, the best way to solve this large program is to discuss individual elements. You mentioned you are very new to programming. As the programs you write become more complex, it's essential to write small working snippets before combining them into a large successful program (or iteratively add snippets). With this being said, it's already apparent you're trying to debug a ~100 line program, and this approach will cost you time in most cases.
The first point to discuss is the general approach you mentioned. If you think about the program, what is the simplest and most repeated step? Obviously this is the kernel/mask step, so we can start from here. When you convolute each pixel, you are performing a similar option, regardless of the position (corner, edge, inside). While there are special steps necessary for these edge cases, they share similar underlying steps. If you try to write code for each of these cases separately, you will have to update the code in multiple (three) places with each adjustment and it will make the whole program more difficult to grasp.
To support my point above, here's what happened when I pasted your code into IntelliJ. This illustrates the (yellow) red flag of using the same code in multiple places:
The concrete way to fix this problem is to combine the three convolve methods into a single one and use if statements for edge-cases as necessary.
Our pseudocode with this change:
convolve(kernel, inputImage)
for each pixel in the image
convolve the single pixel and check edge cases
endfor
end
That seems pretty basic right? If we are able to successfully check edge cases, then this extremely simple logic will work. The reason I left it so general above to show how convolve the single pixel and check edge cases is logically grouped. This means it's a good candidate for extracting a method, which could look like:
private void convolvePixel(int x, int y, int[][] kernel, BufferedImage input, BufferedImage output)
Now to implement our method above, we will need to break it into a few steps, which we may then break into more steps if necessary. We'll need to look at the input image, if possible for each pixel accumulate the values using the kernel, and then set this in the output image. For brevity I will only write pseudocode from here.
convolvePixel(x, y, kernel, input, output)
accumulation = 0
for each row of kernel applicable pixels
for each column of kernel applicable pixels
if this neighboring pixel location is within the image boundaries then
input color = get the color at this neighboring pixel
adjusted value = input color * relative kernel mask value
accumulation += adjusted value
else
//handle this somehow, mentioned below
endif
endfor
endfor
set output pixel as accumulation, assuming this convolution method does not require normalization
end
The pseudocode above is already relatively long. When implementing you could write methods for the if and the else cases, but it you should be fine with this structure.
There are a few ways to handle the edge case of the else above. Your assignment probably specifies a requirement, but the fancy way is to tile around, and pretend like there's another instance of the same image next to this input image. Wikipedia explains three possibilities:
Extend - The nearest border pixels are conceptually extended as far as necessary to provide values for the convolution. Corner pixels are extended in 90° wedges. Other edge pixels are extended in lines.
Wrap - (The method I mentioned) The image is conceptually wrapped (or tiled) and values are taken from the opposite edge or corner.
Crop - Any pixel in the output image which would require values from beyond the edge is skipped. This method can result in the output image being slightly smaller, with the edges having been cropped.
A huge part of becoming a successful programmer is researching on your own. If you read about these methods, work through them on paper, run your convolvePixel method on single pixels, and compare the output to your results by hand, you will find success.
Summary:
Start by cleaning-up your code before anything.
Group the same code into one place.
Hammer out a small chunk (convolving a single pixel). Print out the result and the input values and verify they are correct.
Draw out edge/corner cases.
Read about ways to solve edge cases and decide what fits your needs.
Try implementing the else case through the same form of testing.
Call your convolveImage method with the loop, using the convolvePixel method you know works. Done!
You can look up pseudocode and even specific code to solve the exact problem, so I focused on providing general insight and strategies I have developed through my degree and personal experience. Good luck and please let me know if you want to discuss anything else in the comments below.
Java code for multiple blurs via convolution.

Converting a double scripted array of grayscale int values to a BufferedImage

So after hours of searching I am ready to pull my hair out on this one.
I am doing some research in Computer Vision and am working with grayscale images. I need to end up with an "image" (a double scripted double array) of Sobel filtered double values. My Sobel converter is set up to take in a double scripted int array (int[][]) and go from there.
I am reading in a buffered image and I gather the grayscale int values via a method that I am 99% sure works perfectly (I can present it if need be).
Next I am attempting to convert this matrix of int values to a BufferedImage by the below method:
private BufferedImage getBIFromIntArr(int[][] matrix){
BufferedImage img = new BufferedImage(matrix.length * 4, matrix[0].length, BufferedImage.TYPE_INT_ARGB);
- Gather the pixels in the form of [alpha, r, g, b]
- multiply the size of the array by 4 for the model
int[] pixels = new int[(matrix.length * 4 * matrix[0].length)];
int index = 0;
for (int i = 0; i < matrix.length; i++) {
for (int j = 0; j < matrix[0].length; j++) {
int pixel = matrix[i][j];
pixels[index] = pixel;
index++;
for (int k = 0; k < 3; k++) {
pixels[index] = 0;
index++;
}
}
}
-get the raster
WritableRaster raster = img.getRaster();
-output the amount of pixels and a sample of the array
System.out.println(pixels.length);
for (int i = 0; i < pixels.length; i++) {
System.out.print(pixels[i] + " ");
}
- set the pixels of the raster
raster.setPixels(0, 0, matrix.length, matrix[0].length, pixels);
- paint the image via an external routing to check works (does not)
p.panel.setNewImage(img);
return img;
}
Here is my understanding. The ARGB Type consists of 4 values Alpha, Red, Green, and Blue. I am guessing that setting the alpha values in the new BufferedImage to the greyscale image int values (the matrix values passed in) then this will reproduce the image. Please correct me if I am wrong. So as you can see I create an array of pixels that stores the int values like this: [intValue, 0, 0, 0] repeatedly to try to stay with the 4 value model.
Then I create a writable raster and set the gathered pixels in it using the gathered pixels. The only thing is that I get nothing in the BufferedImage. No error with the code below and Im sure my indeces are correct.
What am I doing wrong? Im sure it is obvious but any help is appreciated because I cant see it. Perhaps my assumption about the model is wrong?
Thanks,
Chronic

Doesn't += return the newly assigned value?

I'm looping over image and summing the values of all pixels. I do this to create an integral image. To keep the last value easily available, I created the_sum variable to which I add values for each pixel.
If you know how integral image works, you know that every pixel in such image contains sum of all pixels before plus it's own value.
Hence:
integral_image[x][y][0] = (the_sum[0]+= (pixel & 0x00FF0000)>>16);
I increase the sum and assign it to current pixel. Netbeans IDE, however, warns me that I'm not reading from the_sum.
Something in the algorithm is broken and I'm not sure what is it. Is my approach wrong or is this a false positive report by NetBeans?
To avoid misunderstanding, this is the whole method:
/* Generate an integral image. Every pixel on such image contains sum of colors or all the
pixels before and itself.
*/
public static double[][][] integralImage(BufferedImage image) {
int w = image.getWidth();
int h = image.getHeight();
double integral_image[][][] = new double[w][h][3];
double the_sum[] = new double[3];
for (int x = 0; x < w; x++) {
for (int y = 0; y < h; y++) {
int pixel = image.getRGB(x, y);
integral_image[x][y][0] = (the_sum[0]+= (pixel & 0x00FF0000)>>16);
integral_image[x][y][1] = (the_sum[1]+= (pixel & 0x0000FF00)>>8);
integral_image[x][y][2] = (the_sum[2]+= pixel & 0x000000FF);
}
}
return integral_image;
}
Yes, += returns the newly assigned value. It is a false positive from netbeans.
At run time, the result of the assignment expression is the value of the variable after the assignment has occurred.
http://docs.oracle.com/javase/specs/jls/se8/html/jls-15.html#jls-15.26

Create image from 2D Color Array

I have an array called image[][] and i want to create a BufferedImage out of this so I can have the player store it in a file.
// Initialize Color[][] however you were already doing so.
Color[][] image;
// Initialize BufferedImage, assuming Color[][] is already properly populated.
BufferedImage bufferedImage = new BufferedImage(image.length, image[0].length,
BufferedImage.TYPE_INT_RGB);
// Set each pixel of the BufferedImage to the color from the Color[][].
for (int x = 0; x < image.length; x++) {
for (int y = 0; y < image[x].length; y++) {
bufferedImage.setRGB(x, y, image[x][y].getRGB());
}
}
This is a straightforward way of creating (and potentially storing) an image, if that's what you're trying to get at. However, this is not efficient by any means. Try it with a larger image and you'll see a noticeable speed difference.

How to change the contrast and brightness of an image stored as pixel values

I have an image that is stored as an array of pixel values. I want to be able to apply a brightness or contrast filter to this image. Is there any simple way, or algorithm, that I can use to achieve this.
Here is my code...
PlanarImage img=JAI.create("fileload","C:\\aimages\\blue_water.jpg");
BufferedImage image = img.getAsBufferedImage();
int w = image.getWidth();
int h = image.getHeight();
int k = 0;
int[] sbins = new int[256];
int[] pixel = new int[3];
Double d = 0.0;
Double d1;
for (int x = 0; x < bi.getWidth(); x++) {
for (int y = 0; y < bi.getHeight(); y++) {
pixel = bi.getRaster().getPixel(x, y, new int[3]);
k = (int) ((0.2125 * pixel[0]) + (0.7154 * pixel[1]) + (0.072 * pixel[2]));
sbins[k]++;
}
}
My suggestion would be to use the built-in methods of Java to adjust the brightness and contrast, rather than trying to adjust the pixel values yourself. It seems pretty easy by doing something like this...
float brightenFactor = 1.2f
PlanarImage img=JAI.create("fileload","C:\\aimages\\blue_water.jpg");
BufferedImage image = img.getAsBufferedImage();
RescaleOp op = new RescaleOp(brightenFactor, 0, null);
image = op.filter(image, image);
The float number is a percentage of the brightness. In my example it would increase the brightness to 120% of the existing value (ie. 20% brighter than the original image)
See this link for a similar question...
Adjust brightness and contrast of BufferedImage in Java
See this link for an example application...
http://www.java2s.com/Code/Java/Advanced-Graphics/BrightnessIncreaseDemo.htm

Categories