High-resolution histogram modification of color images

For color images, histogram modification is usually applied to the quantized luminance component. However, the luminance quantization error can be significantly magnified by the transformation function, leading to distortion in the processed image. The propagation of the quantization error is analyzed theoretically to determine the worst-case error, The relationship between the number of luminance quantization levels and the output quantization error is derived. Experimental results with histogram equalization demonstrate that the use of a higher resolution histogram leads to reduced distortion as well as a“flatter” output histogram.