We propose a new method for image colorization to overcome problems including heavy computation load, poor quality, and excessive human intervention. We choose rectangular blocks from various color images using segmentation and clustering, and calculate color histograms and texture feature points of the blocks. We then build a database by comparing similarity of the blocks based on the color histograms, and find the best matching in terms of the Euclidean distance between the target grayscale blocks and the source color blocks. Having constructed databases of four kinds of image blocks, tree, sky, sand and grass, the target gray image can be colorized using the block database without human intervention.