In order to detect the surface defect on the solder of a Flexible Printed Circuit(FPC), an inspecting technology based on image processing was presented. Firstly, all the defects on the FPC were classified into several defection sorts according to their defection characters. Then, the maximum entropy was used to locate the solder and extract the square and color characters. After estimating the effectness of the Grey Level Co-occurrence Matrix(GLCM) on the quantification for color and structure characters, it was introduced to quantify and extract colorific and structural textures for solders. An analysis on experiments indicates that the defective solder is obviously different from the non-defective solder in several kinds of quantified charaters. On the basis of the result obove,the BP neural network was established and four kinds of characters were selected as the input of neural network. After all neural network weight parameters were adjusted to the optimization through sample training, the performance of the proposed defect detection algorithm was finally evaluated in an on-line testing. Test shows that 50 inspecting targets cost 300 ms, and the inspecting accuracy can reach 94.6%. The experimental result demonstrates that proposed method can detect accurately the solder defect with low false alarms, and the efficiency can satisfy the requirement of defect inspection in online and real time.