English to Spanish translation of signboard images from mobile phone camera

摘要

We propose a whole architecture for English to Spanish translation of the texts present on JPEG natural scene images taken with a mobile phone camera. We detect the text using the frequency information of the DCT coefficients, binarize it using a clustering-based algorithm, recognize it by the use of an OCR algorithm and translate and phonetically transcribe it from English to Spanish. The result is a useful, simple, affordable and robust system that can handle problems like lighting distortion, out of focus, low resolution and foreground-background color variations in a very short amount of processing time, which makes its implementation very attractive on portable devices.