Fuzzy pattern recognition and fuzzy image processing are studied in this dissertation. First, as an aspect of the fuzzy pattern recognition, a neuro-fuzzy classifier is discussed here. Secondly, a fuzzy image processing method based on fuzzy reasoning is used to filter noisy image. The work of this dissertation can be summarized as follows:In the first part of the dissertation, we study NEFCLASS (NEuro Fuzzy CLASSification), a neuro-fuzzy system for the data classification. The membership function should reflect the uncertainty of patterns objectively. Apparently, it's unreasonable to partition the same number fuzzy sets for the different input features in NEFCLASS system. A method is further proposed to choose suitable fuzzy partitions for input features.according to the concept of the fuzzy density. Simulations show that it can improve the correct rate of classification effectively.Detail preserving smoothing of noisy data is a challenging task for any image filtering system. In the second part of the dissertation, for the image corrupted by additive noise, we study FIRE (fuzzy inference ruled by else-action) filter, a fuzzy image filter based on IF-THEN-ELSE fuzzy reasoning.1. A FIRE filter based on hierarchy method is proposed. It can use more information to design fuzzy rules without increasing the implementation complexity overly.2. A new FIRE filter for texture image is presented here. Through integrating pixel patterns based on different spatial supports into a rulebase, a sort of multiresolution image analysis can be performed in a single filter operator. Experimental results show that it is effective for texture image.3. For image corrupted by additive noise, a detection algorithm based on the scale-space theory is proposed. It can classify image pixels into shaped, smooth or textured area. Then the different rulebase can be used for the different area.