The paper consider investigation of the effectiveness of new hybrid algorithm of optimal tone approximation for monochrome multitone images compared to commonly used algorithms. Such algorithms replacing the initial tones of image, which displayed by original palette with some number of tones, on tones from approximating palette that contain less number of tones, but most significant for image displaying.
The hybrid model of optimization consist in combination of heuristic evolutionary-genetic algorithm of tone approximation suboptimization for monochrome images and deterministic algorithm that provides extreme result for the same problem. For estimation the quality of approximation a non-standard criterion of the total module of deviations is used, whose effectiveness compared to traditional square deviation shown in previous investigations. The main goal of evolutionary-genetic algorithm in hybrid model is finding the area of searching optimum according to applied criterion and maximization of reduction that area. Hitting to area of approximating palette structures that are near to optimal, allow to reduce the path of extreme searching for deterministic algorithm, which processing time is much higher, because of using modified, but still brute-force algorithm. For comparison experiments two most common algorithms for tone approximation is used: median cut algorithm and k-means algorithm. The comparison was made based on set of different images. The investigation demonstrated significant advantage of hybrid algorithm against alternative methods and showed increase of approximation quality from 5% to 15%. In addition, different approaches to define the starting position of k-means algorithm were investigated for considered application field, which gives perspectives to make deeper comparative investigations of hybrid algorithm and opportunities to develop new models for tone approximation.