Image Enhancement Techniques Using Digital Image Processing

Image enhancement in digital space is still one of the indispensable pre-processing steps in computer vision workflows,especially when it comes to illumination-uncontrolled image acquisition. In this paper, we conduct an ordered study on spatialand frequency enhancement techniques, arriving at the proposed hybrid enhancement process involving Contrast LimitedAdaptive Histogram Equalization (CLAHE) and illumination correction based on Retinex theory. Our approach makes use ofadaptive histogram clipping and logarithmic illumination splitting in order to overcome local contrast deficiency and globalillumination discrepancy problems at once. Experimental assessment of our method on standard test images proves that our twostage process is more effective in terms of structural integrity than single stage CLAHE or histogram equalization in terms ofentropy, contrast and naturalness preservation indexes. Also, the comparison table of performance features of modernenhancement methods, such as deep-learning-based, variational and swarm methods, is provided. It is shown that properlyarranged classical approaches are still very valuable in constrained environments.