Question

I need to uniformly re-quantize the dynamic range of an image based on the following pixel value conversions:

Pixel Value: Quantized Value
0 - 64     : 31
65 - 128   : 95
129 - 192  : 159
193 - 255  : 223

I want to replace all the pixel values in the above ranges with the quantized values. How can I code this in MATLAB?

Answer 1

One way is to use logical indexing. Given a image matrix img (which could be 2-D grayscale or 3-D RGB), this will replace all your values:

img(img >= 0 & img <= 64) = 31;
img(img >= 65 & img <= 128) = 95;
img(img >= 129 & img <= 192) = 159;
img(img >= 193 & img <= 255) = 223;

Another option is to create a 256-element look-up table and use the values in your image as indices into this table:

lookupTable = [31.*ones(1,65) 95.*ones(1,64) 159.*ones(1,64) 223.*ones(1,63)];
img = uint8(lookupTable(double(img)+1));

Note that with this solution you will have to be mindful of the class of your image matrix. Many images are of class uint8, spanning values 0 to 255. To use these values as an index you have to convert them to a class that can store larger integers (like double) to avoid saturation at the maximum value of 255, then add one since you need an index from 1 to 256. You would then want to convert the resulting image matrix back to class uint8.