Most of the current image database systems depend on visual content to index images, which only provide a partial solution to the image retrieval problem. Natural Scenes are used popularly in our daily lives, which could always cause our strong feelings and senses. This paper discusses how to index natural scenes with season features, one of affective features, to improve the accuracy of image indexing and makes it more convenient to seek them. According to colorful natural scenes? characteristics, it develops sky exclusion plus 1/2 area analysis to extract color features. At the same time, it carries out investigations to collect users? impressions. Then, based on two kinds of features above, it establishes the mapping between color and season features by multiple linear regression, which could be used to index images automatically. Finally, through experiments the mapping is testified valid and correct to forecast and index the season features.