内窥镜头端信噪比判定数据

该数据用于测试摄像头拍摄或者录像时的噪声程度，通过使用带光源内窥镜的摄像头，在Macbeth III标准对色灯箱下，在自带光源的照度lux和色温下，拍摄24色卡卡，使所拍图片占全屏幕的80%左右。提取照片中24色块的19、24灰阶块的亮度值，22灰阶块的噪声值，进行信噪比计算。在内窥镜的头端追求高信噪比，不仅可以显著提高图像质量和诊断准确性，还可以改善手术效果、提高患者安全性、提升设备性能和增强用户体验。因此，高信噪比是内窥镜技术发展的重要方向之一。在开发基于内窥镜图像的AI辅助诊断系统（如识别息肉、肿瘤、出血点）时，模型对颜色非常敏感。本数据可以为AI团队提供色彩准确的“黄金标准”图像，用于模型训练和测试，避免因设备差异导致AI算法失效。同时本数据建立了一个基于客观数据的、可量化的“色彩真实性”标准，可推动行业标准化，推动整个内窥镜行业在色彩质量评估上形成统一标准，从“经验时代”迈向“数字时代”。1、参数说明：S19代表照片中24色块的19灰阶块的信号亮度值。S24代表照片中24色块的24灰阶块的信号亮度值。N22代表照片中24色块的22灰阶块的噪声值。SNR代表信噪比（Signal-to-Noise Ratio)o），是信号与噪声的比值，通常用分贝表示。2、数据处理：将提取到的S19、S24、N22进行如下公式处理得到信噪比的分贝值：SNR=20log10((S19-S24)/N22)。信噪比计算的分贝值越大，代表信号与噪声的比值越高，即信号的质量越好。3、数据输出判断标准：当SNR值较大时，意味着信号的强度远大于噪声的强度，因此信号可以更清晰地被识别和处理。当SNR值较小时，意味着信号的强度接近或小于噪声的强度，这会导致信号难以被识别，从而影响信号的质量和可靠性。设定满足要求的SNR值最低为40，当SNR大于等于40时，输出结果为OK，反之为NG。
单位说明：信号亮度值（S19、S24）：通常用像素值（0到255或0到65535）表示，无单位。噪声值（N22）：通常用像素值的标准差（0到255）表示，无单位。信噪比SNR单位为分贝（dB）。

This dataset is used to test the noise level of a camera during shooting or video recording. Specifically, use a camera equipped with a lighted endoscope to capture the 24-color checker chart in the Macbeth III Standard Color Viewing Booth, under the illuminance (in lux) and color temperature of the booth's built-in light source, ensuring the captured image occupies approximately 80% of the full screen. Extract the signal luminance values of the 19th and 24th grayscale patches, as well as the noise value of the 22nd grayscale patch, from the 24 color patches in the captured photo, then calculate the signal-to-noise ratio (SNR).

Pursuing high signal-to-noise ratio at the distal end of the endoscope can significantly improve image quality and diagnostic accuracy, while also enhancing surgical outcomes, patient safety, device performance, and user experience. Thus, high SNR is one of the key development directions for endoscopy technology.

When developing AI-assisted diagnostic systems based on endoscopic images (such as identifying polyps, tumors, and bleeding spots), the models are highly sensitive to color. This dataset can provide AI teams with color-accurate "gold standard" images for model training and testing, preventing AI algorithms from failing due to device differences. Meanwhile, this dataset establishes a quantifiable "color authenticity" standard based on objective data, which can promote industry standardization, foster a unified standard for color quality assessment across the entire endoscopy industry, and transition the field from the "era of experience" to the "digital era".

1. Parameter Explanation:
S19 represents the signal luminance value of the 19th grayscale patch among the 24 color patches in the captured image.
S24 represents the signal luminance value of the 24th grayscale patch among the 24 color patches in the captured image.
N22 represents the noise value of the 22nd grayscale patch among the 24 color patches in the captured image.
SNR stands for Signal-to-Noise Ratio, which is the ratio of signal to noise, usually expressed in decibels.

2. Data Processing:
The extracted S19, S24, and N22 are processed using the following formula to obtain the SNR value in decibels:
SNR = 20 * log10((S19 - S24) / N22)
A higher decibel value of the calculated SNR indicates a higher ratio of signal to noise, meaning better signal quality.

3. Data Output Judgment Criteria:
A larger SNR value means the signal intensity is much greater than the noise intensity, so the signal can be recognized and processed more clearly. A smaller SNR value means the signal intensity is close to or less than the noise intensity, which will make the signal difficult to recognize, thereby affecting the quality and reliability of the signal. The minimum required SNR value is set to 40. When the SNR is greater than or equal to 40, the output result is OK; otherwise, it is NG.

Unit Explanation:
Signal luminance values (S19, S24): Usually expressed as pixel values (0 to 255 or 0 to 65535), with no unit.
Noise value (N22): Usually expressed as the standard deviation of pixel values (0 to 255), with no unit.
The unit of SNR is decibel (dB).

搜集汇总

背景与挑战

背景概述

该数据集由浙江优亿医疗器械股份有限公司提供，包含1002条企业数据，以xlsx格式存储，用于内窥镜头端信噪比判定。数据集通过提取24色卡灰阶块的亮度值和噪声值，计算信噪比（SNR），并根据阈值（SNR≥40为OK，否则为NG）评估图像质量，旨在提升内窥镜诊断准确性和支持AI辅助系统开发，推动行业标准化进程。