I'd like to sum my 4X4 block. Suppose I have an image and will divide it into 4X4 blocks. Then afterward I'd like to determine the sum of each block using cvIntegral. How can I cope this?
Here is my basic program in order to calculate integral image value of whole image:
float s = 0.0f;
//Read in the image
IplImage* hImage = cvLoadImage("bayer-image.jpg",0);
UINT width = hImage->width; UINT height = hImage->height;
CvMat* sum = cvCreateMat(height + 1, width + 1, CV_32SC1);
CvMat* sqsum = cvCreateMat(height + 1, width + 1, CV_64FC1);
cvIntegral(hImage, sum, sqsum);
cvReleaseImage(&hImage);
cvReleaseMat(&sum);
cvReleaseMat(&sqsum);
What should I do next?
Really thanks in advance.
Check this out
http://en.wikipedia.org/wiki/Summed_area_table
As an example, the block defined by the corners (1, 1) and (4, 4) has the area
a1 = integral(0,0)+integral(4,4)-integral(4,0)-integral(0,4);
Related
https://stackoverflow.com/a/2574798/159072
public static Bitmap BitmapTo1Bpp(Bitmap img)
{
int w = img.Width;
int h = img.Height;
//
Bitmap bmp = new Bitmap(w, h, PixelFormat.Format1bppIndexed);
BitmapData data = bmp.LockBits(new Rectangle(0, 0, w, h), ImageLockMode.ReadWrite, PixelFormat.Format1bppIndexed);
Why this addition and division?
byte[] scan = new byte[(w + 7) / 8];
for (int y = 0; y < h; y++)
{
for (int x = 0; x < w; x++)
{////Why this condition check?
if (x % 8 == 0)
//Why divide by 8?
scan[x / 8] = 0;
Color c = img.GetPixel(x, y);
//Why this condition check?
if (c.GetBrightness() >= 0.5)
{
// What is going on here?
scan[x / 8] |= (byte)(0x80 >> (x % 8));
}
}
// Why Martial.Copy() called here?
Marshal.Copy(scan, 0, (IntPtr)((long)data.Scan0 + data.Stride * y), scan.Length);
}
bmp.UnlockBits(data);
return bmp;
}
The code uses some basic bit-hacking techniques, required because it needs to set bits and the minimum storage element you can address in C# is a byte. I intentionally avoided using the BitArray class.
int w = img.Width;
I copy the Width and Height properties of the bitmap into a local variable to speed up the code, the properties are too expensive. Keep in mind that w are the number of pixels across the bitmap, it represents the number of bits in the final image.
byte[] scan = new byte[(w + 7) / 8];
The scan variable stores the pixels in one scan line of the bitmap. The 1bpp format uses 1 bit per pixel so the total number of bytes in a scan line is w / 8. I add 7 to ensure the value is rounded up, necessary because integer division always truncates. w = 1..7 requires 1 byte, w = 8..15 requires 2 bytes, etcetera.
if (x % 8 == 0) scan[x / 8] = 0;
The x % 8 expression represents the bit number, x / 8 is the byte number. This code sets all the pixels to Black when it progresses to the next byte in the scan line. Another way to do it would be re-allocating the byte[] in the outer loop or resetting it back to 0 with a for-loop.
if (c.GetBrightness() >= 0.5)
The pixel should be set to White when the source pixel is bright enough. Otherwise it leaves it at Black. Using Color.Brightness is a simple way to avoid dealing with the human eye's non-linear perception of brightness (luminance ~= 0.299 * red + 0.587 * green + 0.114 * blue).
scan[x / 8] |= (byte)(0x80 >> (x % 8));
Sets a bit to White in the scan line. As noted x % 8 is the bit number, it shifts 0x80 to the right by the bit number, they are stored in reverse order in this pixel format.
Overview
In my app (which is a game), I make use of the batching of items to reduce the number of draw calls. So, I'll, create for example, a Java object called platforms which is for all the platforms in the game. All the enemies are batched together as are all collectible items etc....
This works really well. At present I am able to size and position the individual items in a batch independently of each other however, I've come to the point where I really need to change the opacity of individual items also. Currently, I can change only the opacity of the entire batch.
Batching
I am uploading the vertices for all items within the batch that are to be displayed (I can turn individual items off if I don't want them to be drawn), and then once they are all done, I simply draw them in one call.
The following is an idea of what I'm doing - I realise this may not compile, it is just to give an idea for the purpose of the question.
public void draw(){
//Upload vertices
for (count = 0;count<numOfSpritesInBatch;count+=1){
vertices[x] = xLeft;
vertices[(x+1)] = yPTop;
vertices[(x+2)] = 0;
vertices[(x+3)] = textureLeft;
vertices[(x+4)] = 0;
vertices[(x+5)] = xPRight;
vertices[(x+6)] = yTop;
vertices[(x+7)] = 0;
vertices[(x+8)] = textureRight;
vertices[x+9] = 0;
vertices[x+10] = xLeft;
vertices[x+11] = yBottom;
vertices[x+12] = 0;
vertices[x+13] = textureLeft;
vertices[x+14] = 1;
vertices[x+15] = xRight;
vertices[x+16] = yTop;
vertices[x+17] = 0;
vertices[x+18] = textureRight;
vertices[x+19] = 0;
vertices[x+20] = xLeft;
vertices[x+21] = yBottom;
vertices[x+22] = 0;
vertices[x+23] = textureLeft;
vertices[x+24] = 1;
vertices[x+25] = xRight;
vertices[x+26] = yBottom;
vertices[x+27] = 0;
vertices[x+28] = textureRight;
vertices[x+29] = 1;
x+=30;
}
vertexBuf.rewind();
vertexBuf.put(vertices).position(0);
GLES20.glBindTexture(GLES20.GL_TEXTURE_2D, texID);
GLES20.glUseProgram(iProgId);
Matrix.multiplyMM(mvpMatrix2, 0, mvpMatrix, 0, mRotationMatrix, 0);
mMVPMatrixHandle = GLES20.glGetUniformLocation(iProgId, "uMVPMatrix");
GLES20.glUniformMatrix4fv(mMVPMatrixHandle, 1, false, mvpMatrix2, 0);
vertexBuf.position(0);
GLES20.glVertexAttribPointer(iPosition, 3, GLES20.GL_FLOAT, false, 5 * 4, vertexBuf);
GLES20.glEnableVertexAttribArray(iPosition);
vertexBuf.position(3);
GLES20.glVertexAttribPointer(iTexCoords, 2, GLES20.GL_FLOAT, false, 5 * 4, vertexBuf);
GLES20.glEnableVertexAttribArray(iTexCoords);
//Enable Alpha blending and set blending function
GLES20.glEnable(GLES20.GL_BLEND);
GLES20.glBlendFunc(GLES20.GL_ONE, GLES20.GL_ONE_MINUS_SRC_ALPHA);
//Draw it
GLES20.glDrawArrays(GLES20.GL_TRIANGLES, 0, 6 * numOfSpritesInBatch);
//Disable Alpha blending
GLES20.glDisable(GLES20.GL_BLEND);
}
Shaders
String strVShader =
"uniform mat4 uMVPMatrix;" +
"attribute vec4 a_position;\n"+
"attribute vec2 a_texCoords;" +
"varying vec2 v_texCoords;" +
"void main()\n" +
"{\n" +
"gl_Position = uMVPMatrix * a_position;\n"+
"v_texCoords = a_texCoords;" +
"}";
String strFShader =
"precision mediump float;" +
"uniform float opValue;"+
"varying vec2 v_texCoords;" +
"uniform sampler2D u_baseMap;" +
"void main()" +
"{" +
"gl_FragColor = texture2D(u_baseMap, v_texCoords);" +
"gl_FragColor *= opValue;"+
"}";
Currently, I have a method in my Sprite class that allows me to change the opacty. For example, something like this:
spriteBatch.setOpacity(0.5f); //Half opacity
This works, but changes the whole batch - not what I'm after.
Application
I need this because I want to draw small indicators when the player destroys an enemy - which show the score obtained from that action. (The type of thing that happens in many games) - I want these little 'score indicators' to fade out once they appear. All the indicators would of course be created as a batch so they can all be drawn with one draw call.
The only other alternatives are:
Create 10 textures at varying levels of opacity and just switch between them to create the fading effect. Not really an option as way too wasteful.
Create each of these objects separately and draw each with their own draw call. Would work, but with a max of 10 of these objects on-screen, I could potentially be drawing using 10 draw calls just for these items - while the game as a whole currently only uses about 20 draw calls to draw a hundreds of items.
I need to look at future uses of this too in particle systems etc.... so I would really like to try to figure out how to do this (be able to adjust each item's opacity separately). If I need to do this in the shader, I would be grateful if you could show how this works. Alternatively, is it possible to do this outside of the shader?
Surely this can be done in some way or another? All suggestions welcome....
The most direct way of achieving this is to use a vertex attribute for the opacity value, instead of a uniform. This will allow you to set the opacity per vertex, without increasing the number of draw calls.
To implement this, you can follow the pattern you already use for the texture coordinates. They are passed into the vertex shader as an attribute, and then handed off to the fragment shader as a varying variable.
So in the vertex shader, you add:
...
attribute float a_opValue;
varying float v_opValue;
...
v_opValue = a_opValue;
...
In the fragment shader, you remove the uniform declaration for opValue, and replace it with:
varying float v_opValue;
...
gl_FragColor *= v_opValue;
...
In the Java code, you extend the vertex data with an additional value for the opacity, to use 6 values per vertex (3 position, 2 texture coordinates, 1 opacity), and update the state setup accordingly:
vertexBuf.position(0);
GLES20.glVertexAttribPointer(iPosition, 3, GLES20.GL_FLOAT, false, 6 * 4, vertexBuf);
GLES20.glEnableVertexAttribArray(iPosition);
vertexBuf.position(3);
GLES20.glVertexAttribPointer(iTexCoords, 2, GLES20.GL_FLOAT, false, 6 * 4, vertexBuf);
GLES20.glEnableVertexAttribArray(iTexCoords);
vertexBuf.position(5);
GLES20.glVertexAttribPointer(iOpValue, 1, GLES20.GL_FLOAT, false, 6 * 4, vertexBuf);
GLES20.glEnableVertexAttribArray(iOpValue);
I'd like to determine the mean block of my image using histogram. Let's say my image has 64 by 64 dimension, I need to divide it into 4 by 4 block then determine each block mean (in other word now I will have 4 blocks).
Using opencv, How do I can utilize my IplImage to determine block mean using histogram bins?
The code below is opencv histogram in order to determine whole image mean:
int i, hist_size = 256;
float max_value,min_value;
float min_idx,max_idx;
float bin_w;
float mean =0, low_mean =0, high_mean =0, variance =0;
float range_0[]={0,256};
float *ranges[]={range_0};
IplImage* im = cvLoadImage("killerbee.jpg");
//Create a single planed image of the same size as the original
IplImage* grayImage = cvCreateImage(cvSize(im->width,im->height),IPL_DEPTH_8U, 1);
//convert the original image to gray
cvCvtColor(im, grayImage, CV_BGR2GRAY);
/* Remark this, since wanna evaluate whole area.
//create a rectangular area to evaluate
CvRect rect = cvRect(0, 0, 500, 600 );
//apply the rectangle to the image and establish a region of interest
cvSetImageROI(grayImage, rect);
End remark*/
//create an image to hold the histogram
IplImage* histImage = cvCreateImage(cvSize(320,200), 8, 1);
//create a histogram to store the information from the image
CvHistogram* hist = cvCreateHist(1, &hist_size, CV_HIST_ARRAY, ranges, 1);
//calculate the histogram and apply to hist
cvCalcHist( &grayImage, hist, 0, NULL );
//grab the min and max values and their indeces
cvGetMinMaxHistValue( hist, &min_value, &max_value, 0, 0);
//scale the bin values so that they will fit in the image representation
cvScale( hist->bins, hist->bins, ((double)histImage->height)/max_value, 0 );
//set all histogram values to 255
cvSet( histImage, cvScalarAll(255), 0 );
//create a factor for scaling along the width
bin_w = cvRound((double)histImage->width/hist_size);
for( i = 0; i < hist_size; i++ ) {
//draw the histogram data onto the histogram image
cvRectangle( histImage, cvPoint(i*bin_w, histImage->height),cvPoint((i+1)*bin_w,histImage->height - cvRound(cvGetReal1D(hist->bins,i))),cvScalarAll(0), -1, 8, 0 );
//get the value at the current histogram bucket
float* bins = cvGetHistValue_1D(hist,i);
//increment the mean value
mean += bins[0];
}
//finish mean calculation
mean /= hist_size;
//display mean value onto output window
cout<<"MEAN VALUE of THIS IMAGE : "<<mean<<"\n";
//go back through now that mean has been calculated in order to calculate variance
for( i = 0; i < hist_size; i++ ) {
float* bins = cvGetHistValue_1D(hist,i);
variance += pow((bins[0] - mean),2);
}
//finish variance calculation
variance /= hist_size;
cvNamedWindow("Original", 0);
cvShowImage("Original", im );
cvNamedWindow("Gray", 0);
cvShowImage("Gray", grayImage );
cvNamedWindow("Histogram", 0);
cvShowImage("Histogram", histImage );
//hold the images until a key is pressed
cvWaitKey(0);
//clean up images
cvReleaseImage(&histImage);
cvReleaseImage(&grayImage);
cvReleaseImage(&im);
//remove windows
cvDestroyWindow("Original");
cvDestroyWindow("Gray");
cvDestroyWindow("Histogram");
Really thanks in advance.
You can do that by histograms, but a much more effective way to do it is an integral image, which does almost what you want.
Read here http://en.wikipedia.org/wiki/Summed_area_table and then use it to calculate the sum of all the pixels in every block. Then divide by the number of pixels in each block (4x4=16). Isn't it nice?
OpenCV has a function to calculate the integral image, with the difficult name cv::integral()
And an even easier way to do it is the humble resize().
Call resize(image64_64, image_16_16, Size(16, 16), INTER_AREA), and the result will be a smaller image whose pixel values have exactly the values you're looking for. Isn't it great?
Just do not forget the INTER_AREA flag. It determines the correct algorithm to be used.
Is there any problem in this code where I am just trying to subtract the pixel values of images through direct access of pixels ..... Am assuming that the images are of same height and width ... Whenever I run the program I am getting completely black picture.....
IplImage * img3 = cvCreateImage(cvSize(img1->height,img1->width),IPL_DEPTH_32F,3);
// img2 and img1 both are IplImage pointers
cvZero(img3);
long value;
for ( int row = 0 ; row < img2->height * img2->width ; row ++ ){
value = &((uchar*)(img1->imageData))[row] - &((uchar*)(img2->imageData))[row] ;
img3->imageData[row] = value;
1) img2->height * img2->width calculate as constant before loop
2)
I dont understand this line
&((uchar*)(img1->imageData))[row] - &((uchar*)(img2->imageData))[row] - are you subtracting pointer from another pointer? Why?
value = img1->imageData[row] - img2->imageData[row]; should do the trick
3) you can not subtract RGB values by subtracting pixel values (if that is your goal)
4) if img3->imageData is *char, then you should multiply row * 4.
I want to fill the intersection of two(or more filled) rectangles with the average color. I have the colors of each rectangle stored as unsigned ints. How can I get the average color?
Thank you for you help!
Technically, you might be running on a color-map device, which means you need to go through X11 color management for all of this. You need to query the XColor for your two input colors, compute the average, then look up the closest representable color:
// Query XColor for both input colors
XColor xcol1, xcol2, outcol;
xcol1.pixel = color1;
xcol2.pixel = color2;
XQueryColor(display, colormap, &xcol1);
XQueryColor(display, colormap, &xcol2);
// Average red/green/blue and look up nearest representable color
outcol.red = (xcol1.red + xcol2.red) / 2;
outcol.green = (xcol1.green + xcol2.green) / 2;
outcol.blue = (xcol1.blue + xcol2.blue) / 2;
XAllocColor(display, colormap, &outcol);
// outcol.pixel is now the color to use
On a paletted device, you also need to free the color afterwards etc. - it's a mess, basically.
But in all likelihood you're on a 32-bit truecolor device, which means the integer is just a bitfield of r, g, b and a (not necessarily in that order). You can compute their average like this:
UInt out_color = 0;
for (int i=0; i < 4; i++) {
// Extract channel i from both input colors
UInt in1 = (color1 >> (i*8)) & 0xff;
UInt in2 = (color2 >> (i*8)) & 0xff;
// Compute the average and or it into the output color
out_color |= ((in1 + in2) / 2) << (i*8);
}
Color color1 = Color.FromArgb(UInt1);
Color color2 = Color.FromArgb(UInt2);
Color averageColor = Color.FromArgb(255,(color1.R + color2.R)/2,(color1.G + color2.G)/2,(color1.B + color2.B)/2);
This is assuming that you need a fully opaque average color.