The advancement of autonomous driving (AD) technology to SAE level 4 autonomy has increased the demand for operation of highly safe systems in autonomous vehicles (AVs). A critical aspect of this autonomy is a reliable lane detection system, crucial for precise lane keeping and maneuvering. However, existing systems often depend on a single camera, leading to safety concerns arising from potential hardware failures or communication issues. To address this, there is a pressing need to enhance the reliability of lane detection systems by identifying and rectifying failures within the single-camera perception system. This thesis proposes a real-time dual image cropping algorithm designed to evaluate image overlap between two independently operated cameras, each providing input for lane detection algorithms running in distinct, failure-independent hardware channels. Additionally, the algorithm utilizes a redundancy evaluator algorithm to cross-reference the outputs of each lane detection algorithm, effectively validating redundancy between channels and identifying potential calculation errors resulting from hardware and communication faults. This innovative approach represents a novel application of the ISO26262 Automotive Safety Integrity Level (ASIL) decomposition principle within a perception system. The primary objective is to elevate safety standards to ASIL-D in AVs operating at Level-4 autonomy by developing a high-accuracy image cropping algorithm that operates within computational constraints and facilitates reliable redundant perception channels. The evaluation of the redundancy between the channels is analyzed using a quantitative comparison of the performance of the two lane detection algorithms by assessing the mean absolute error between their output values (distance to the left and right lane markings) across multiple image samples. The system was able to produce promising results with mean absolute error of just 3cm, which is an indication that the introduced system of the dual cropping algorithm can be used to produce highly similar images, enabling redundant channels and increasing the safety of an AD system. The algorithm's real-time performance is also measured through metrics such as publishing rate, processing time, and latency among the cropping image publishers. The average processing time for cropping was approximately 184 milliseconds, with a mere 5 milliseconds of latency between publishers operating at a frequency of 3 frames per second (FPS). These findings robustly affirm the algorithm's suitability for real-time applications in autonomous driving.