With the popularity of digital still cameras (DSCs), improvement of image quality in dark areas of the image is needed. There are two typical techniques for this: gamma correction and histogram equalization. However, such techniques are not always sufficient to improve scene detail in dark areas. Recently, examinations of Retinex theory taking into account the human vision model proposed by Land are being given attention. This algorithm, which utilizes spatial information between surrounding pixels, gives good image detail. Single-scale Retinex (SSR) and Multi-scale Retinex (MSR) are typical Retinex algorithms. However, they raise several practical use issues concerning color images reproduced by printers. In order to address such issues, Adaptive Multi-scale Retinex (AMSR) synthesis of images originally processed by MSR to the original image has been proposed. This algorithm consists of two processes, linear computation and synthesis of both the original image and the image processed by MSR. AMSR can be compared to histogram equalization and MSR for DSC images. AMSR shows that visibility in dark areas can be improved compared to other techniques. Moreover, NEW-AMSR has the following two features: suppression of chromatic unbalance in R, G and B channels, and a high speed processing technique to compute a weighted average.