10xEngineers

Our ISP team recently updated the Infinite-ISP pipeline by adding a Sharpening block in the Infinite-ISP FPGA Binaries v1.4 Release. The Infinite-ISP Reference Design on AMD Xilinx Kria KV260 Kit supports the Raspberry Pi Camera Module v2, providing users with full sensor control and producing crisp colour rendering with Sony’s IMX219 image sensor. 👉 Visit the Infinite-ISP Project on our GitHub and try out our latest binaries today: https://lnkd.in/dtrsXAcN Need help in deploying our open-source hardware Image Signal Processor on FPGA? 👉 Check out our tutorial here: https://lnkd.in/dMctDWau 10xEngineers

Keeping the momentum alive—our second paper hits the mark! We are proud to share that our second research paper, "𝐈𝐧𝐟𝐢𝐧𝐢𝐭𝐞-𝐈𝐒𝐏: 𝐀𝐧 𝐎𝐩𝐞𝐧 𝐒𝐨𝐮𝐫𝐜𝐞 𝐇𝐚𝐫𝐝𝐰𝐚𝐫𝐞 𝐈𝐦𝐚𝐠𝐞 𝐒𝐢𝐠𝐧𝐚𝐥 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐨𝐫 𝐏𝐥𝐚𝐭𝐟𝐨𝐫𝐦 𝐟𝐨𝐫 𝐀𝐥𝐥 𝐈𝐦𝐚𝐠𝐢𝐧𝐠 𝐍𝐞𝐞𝐝𝐬" has been accepted at Electronic Imaging 2025! This marks a major milestone for us at 10xEngineers as we continue to make waves in the imaging tech world. Two accepted papers in a prestigious conference is no small feat, and we couldn’t be more excited to contribute to the future of open-source imaging innovation. 𝐀𝐛𝐨𝐮𝐭 𝐈𝐧𝐟𝐢𝐧𝐢𝐭𝐞-𝐈𝐒𝐏: Infinite-ISP is a comprehensive open-source hardware Image Signal Processor (ISP) development suite. It offers a full-stack solution encompassing algorithm research, RTL development, and FPGA/ASIC implementation, complemented by an ISP Tuning Tool. 👉 Click here to know more about us: https://lnkd.in/dw4y5QuK 10xEngineers Society for Imaging Science and Technology (IS&T) #EI2025

10xEngineers HDR ISP, The right choice for your next gen camera modules. Our HDR ISP outshines its competition, offering peak image quality with minimal footprint and providing artifact-free HDR processing even in low-light conditions! 10xEngineers' proprietary image signal processing algorithm chain makes true scene dynamic range available to your customers. 👉 Click here to visit our website: https://lnkd.in/d_tqAK2e

🚨 Reminder: Webinar Tomorrow!🚨 We’re excited to host a webinar on "Auto Camera Tuning for Vision Models" tomorrow, where Khurram Mazher will explore the Auto Camera Tuning (ACT) paradigm. This innovative approach enables tuning of ISPs specifically for deep learning vision models and applications, rather than just optimizing visual image quality for human perception. Khurram will demonstrate how ISPs can be optimized to enhance machine vision metrics such as mean Average Precision (mAP) and mean Intersection over Union (mIoU), achieving performance improvements of 5-15% for pretrained models and 10-25% with joint ISP and model tuning. 📅 Your participation is highly recommended! Don’t miss this opportunity to dive into advancements in vision model optimization. ⌚Time: Dec 3, 2024 03:00 PM Islamabad, Karachi, Tashkent 🧑🤝🧑Join Zoom Link: https://lnkd.in/dh8mUgAa Meeting ID: 867 5062 3995 Passcode: 712433

Our Senior Engineer, Khurram Mazher, will be presenting an exclusive webinar on Auto Camera Tuning for Vision Models! Join us as Khurram shares their insights into the latest techniques driving innovation in computer vision, unlocking new levels of performance and precision for AI applications. This is a must-attend event for professionals and enthusiasts eager to stay ahead in the field of AI and vision technologies. 📅 Dec 3rd, 2024 & 3:00 PM 👉 Join Zoom Meeting https://lnkd.in/dh8mUgAa Meeting ID: 867 5062 3995 Passcode: 712433 Don’t miss out on this opportunity to learn from one of the industry’s finest! 💡Mark your calendars today. Visit our website to know more about us: https://10xengineers.ai/

As we work on enhancing our HDR ISP pipeline, one challenge continues to stand out: the scarcity of accessible, open-source benchmarks for HDR image processing. While some tools are available, many come with complex setup requirements, limited documentation, or platform-specific constraints, making it difficult to use them effectively for consistent benchmarking. Without widely-adopted, easy-to-use open-source HDR processing pipelines, comparing HDR image quality in a meaningful, standardized way remains a challenge. Commercial tools like Capture One and HDR Express allow for subjective assessments of HDR quality, but they fall short in providing the flexibility and transparency we’d ideally want for benchmarking. Moreover, these proprietary options don’t provide access to the underlying processing algorithms, making it hard to tailor or optimize evaluations for specialized needs. Has anyone come across reliable, open-source HDR pipelines that are easy to run on Linux/Windows? Or, if you’re facing similar challenges, how are you benchmarking your HDR results? What’s the “golden standard” for subjective HDR image quality in fields like automotive and surveillance? Comment below to exchange insights about subjective standards and tools for HDR image processing pipelines! 👉 Visit our website here: https://lnkd.in/d_tqAK2e 10xEngineers

The Sony IMX219 image sensor is present in the Raspberry Pi Camera Module v2.x. It provides many on-sensor Image Signal Processing (ISP) operations such as Digital Gain setting, Black Level Correction, Defect Pixel Correction, etc. This post will look into the technical aspects of the on-sensor black-level clamp function. In image sensors, due to the presence of dark currents in the inherent pixel grid circuitry, a DC offset is introduced into the ADC-sampled signal, which is called black level offset. Due to varying operating temperatures, the inherent dark currents in the pixels also vary, leading to variation in the offset in the digital values. A change in ambient temperature can also contribute to the change in black level. Tackling these varying black levels in external ISPs for varying ranges of temperatures and analog gains is a challenging task. To address this issue, IMX219 includes a closed-loop black-level clamping function that clamps the black level across the pixels to a constant value of 64 for RAW10. This is achieved by subtracting the error from the analog signal in a negative feedback system similar to the one shown in the image. This introduces a uniform black level for pixel grid across a range of temperature and analog gain, simplifying the work for external ISPs. Tackling this uniform black level is easier and applying the relevant correction in the external ISP is relatively easier compared to the black levels in other image sensors that may vary with temperature and sensor gain. It also yields better image quality than the other sensor types. For more insights into how Black Level Clamping and Correction affects the overall quality of your image, check out this video demonstration: https://lnkd.in/dK49ggfu 👉Visit our website to drop your query: https://10xengineers.ai/ 10xEngineers #BlackLevelCorrection