Research On Grayscale Colorization Method Based On Clustering

Color is the most important factor in human visual perception.Colorization is a key technology and application scenario for information collection,identification and application in the digital age.Relying on digital technology and colorization to transform black and white images made by early monochrome photography into color effects.Real color images are becoming a trend to restore the colorful world.Studying the colorization of grayscale images has become the theory in the field of digital image processing.It has great research value and application significance for the transformation and innovation of remote sensing satellite,medicine,art and life.In this paper,based on the comprehensive analysis of the traditional color-based gray-scale colorization method,a color-based colorization method based on clustering is proposed for the problem that the method is easy to produce color transmission error and the edge region is not smooth.This method is aimed at the key technology of colorization.In color space conversion and image segmentation,an improved GMM clustering method is designed and proposed.By introducing the penalty likelihood function and spatial smoothing parameters,the EM algorithm is implemented on GMM.The optimization of the model's parameter-seeking process yields the number of sub-blocks for optimal cluster segmentation.In the correspondence between sub-blocks,four characteristics of brightness,standard deviation,Gabor texture and Daisy are selected.By combining these characteristics,a fast cascade matching method with better efficiency is formed.In the color transfer,the method based on brightness and texture features is proposed,which effectively solves the problem of low precision based on the traditional color-based color transfer method.In the evaluation of colorization results,an image quality evaluation method based on subjective and objective combination of human visual perception and parameter index is proposed.The rationality and effectiveness of the proposed method are verified by comparison experiments,and the expected results are achieved.