An energy efficient multi-scale retinex algorithm based on the ratio of saturated regions

This paper proposes an image enhancement technique that adjusts the luminance of an input image depending on the image characteristics. First, the proposed method classifies the input image into three types based on the ratio of saturated region. After the classification, the enhancement process is adaptively performed depending on the classified image type to improve the image quality and reduce the power consumption. In the enhancement process, the proposed method reduces computational complexity for the low power consumption when calculating the illumination. Thus, the proposed method has an advantage of effectively reducing the power consumption as much as possible, while improving the output image quality. The experimental results showed that the proposed method successfully reduced the power consumption by up to 90.4% while increasing the enhancement performance measure by up to 73% and obtaining the highest subjective scores compared to the benchmark methods.