To address the data island issue caused by outdated or unconnected relay protection devices in power grid fault analysis, this paper proposes an automated parsing and quantification framework based on computer vision. Currently, the analysis of such image-based reports relies heavily on manual interpretation, which is inefficient and error-prone. This limitation not only delays fault response but also increases the risk of secondary accidents due to misjudgments. The proposed framework integrates Optical Character Recognition (OCR) and a waveform quantification algorithm, enabling simultaneous extraction of discrete textual information and continuous waveform data from a single report image. The core algorithm robustly measures the width of static scale markers and the maximum dynamic width of transient waveforms by directly analyzing grayscale pixel features in screenshots. A real-world fault case from a CSC-103 relay device demonstrates that the framework can efficiently and accurately convert unstructured image data into structured numerical information. The extracted results are consistent with protection logic. This study offers a low-cost, high-robustness digital solution to empower existing power infrastructure, enhancing the accuracy and efficiency of fault diagnostics in practical engineering scenarios.