e-KTP (Kartu Tanda Penduduk Elektronik) is a national identity card for Indonesian has been widely used not only as an identification but also used in various aspects of life such as for administration, finance-related matters, etc. Detecting information from the ID card field can be done using previous methods with segmentation and text extraction. However, these methods need a proper high-quality captured image. In fact, most of the ID card images are taken using mobile phones' cameras, which do not really produce good quality images all the time. Image enhancement using GAN (Generative Adversarial Network) was also proposed as a method of increasing the accuracy of OCR (Optical Character Recognition). But in previous studies, this method was only limited to document text images with a white background. To overcome these problems, we propose a new pre-processing approach consists of DeblurGAN (Generative Adversarial Network for deblurring image), shadow removal, and binarization to pre-process the image for Tesseract-OCR. We also propose a simple post-processing method for Tesseract-OCR output to extract key-value pairs for each field in e-KTP. By using our approach, we have achieved an average Character Error Rate of 18.82% which is better compared to without pre-processing which is 38.13%.