A method for operating optical sensors, particularly code readers, which have a light receptor that receives light reflected and/or returned from an object. A picture is generated from the received light and the content of the picture, such as a code, is read. At least some pictures are compressed with a picture compression algorithm of the sensor. To improve the operation of such optical sensors, especially code readers, and to improve the analysis of error detection from knowing the causes for errors, only pictures which have error information are compressed and stored. The code reader is constructed to perform these functions.