In order to balance image quality and power consumption， a local dimming algorithm based on image local features was proposed. Firstly， the image features of the corresponding display area were extracted and classified； then， the initial backlight value of the region was obtained based on the improved error correction method. The brightness dimming factor was adjusted according to the image detail characteristics to finally determine the output backlight value. Finally， the input display image was separated by Gaussian filter and the pixel compensation based on S-curve was carried out according to the different detail content. The experimental results showed that compared with the display without dimming technology， the backlight power consumption was reduced by 38.36%.Compared with the error correction algorithm， the average peak signal to noise ratio （PSNR） was increased by 7.83 dB and the average information entropy was increased by 6.32%.The proposed algorithm could greatly reduce the display power consumption and improve the static contrast， as well as retain more details of the original images.