Due to the influence of materials and processes， infrared images generally suffer from the problem of blind pixels. For infrared images from new Blocked Impurity Band （BIB） detectors， there are still issues such as limited dynamic range and significant non-uniformity. Conventional blind pixel detection and compensation methods are not fully applicable to BIB detector images. To address this issue， this paper enumerates the pros and cons of common methods for detecting and compensating blind pixels in infrared images and conducts experimental processing on actual measured BIB images one by one. However， the results indicate that the distribution of blind pixels is highly non-uniform， with a relatively high proportion of clustered blind pixels. Therefore， this paper proposes an improved blind pixel detection method and a blind pixel compensation method， and implements the algorithms using an FPGA-based hardware system platform. The analysis shows that the uniformity of blind pixel distribution and the proportion of clustered blind pixels have been optimized after the improvement， leading to a tangible enhancement in their economic viability for application.