normalized Image Coordinate

本文介绍了一种用于统一处理不同模块间图像坐标的方法。通过将像素坐标转换为标准化坐标,使得坐标范围固定在[-1,1]之间,便于跨分辨率操作。标准化后的坐标以图像中心为原点,水平方向向右、垂直方向向下递增。

摘要生成于 C知道 ,由 DeepSeek-R1 满血版支持, 前往体验 >

To provide a uniform treatment of image coordinates between modules, independently of image size / resolution (different modules may use different image resolutions), it is convenient to adopt the following standard:

  • Let the image have size (w, h), where w is the image width and h is the image height.
  • Let the image coordinates be (u,v), in pixel units, where u is the horizontal coordinate, starting at the top-left corner and growing to the right, and v is the vertical image coordinate, starting at the top-left corner and growing to the bottom.
  • The normalized image coordinates are defined as:

x = 2*u/w - 1

y = 2*v/h - 1

Thus, the normalized image coordinates (both x and y) are in the range [-1,1], being 0 the center of the image, -1 the left/top boundaries, and 1 the right/bottom boundaries.

The x coordinate starts at the image center and grows to the right. The y coordinate starts at the image center and grows to the bottom.

To go back to image pixels, the following expressions should be used:

u = w*(x+1)/2

v = h*(y+1)/2

{ "dataset_name": "tnbc", "dataset_type": "segmentation", "total_samples": 50, "image_dimensions": [ 512, 512 ], "channels": 3, "task_type": "nuclei_segmentation", "file_format": "h5", "creation_date": "2025-07-10T15:20:01.296151", "processing_info": { "image_preprocessing": "normalized_to_0_1", "mask_encoding": "instance_segmentation", "coordinate_system": "xy_origin_top_left" }, "statistics": { "total_nuclei": 4027, "avg_nuclei_per_image": 80.54, "min_nuclei_per_image": 3, "max_nuclei_per_image": 295, "image_size_distribution": { "512x512": 50 }, "mask_area_statistics": { "avg_area": 406.10504097342937, "min_area": 10, "max_area": 3205, "total_area": 1635385 } }, "file_info": { "h5_file": "tnbc.h5", "json_file": "tnbc.json", "size_mb": 38.562448501586914 }, "dataset_source": "https://zenodo.org/records/2579118", "dataset_description": "Triple negative breast cancer nuclei segmentation dataset", "patient_count": 11, "annotation_format": "binary_ground_truth", "image_format": "png", "cancer_type": "triple_negative_breast_cancer", "staining": "H&E", "magnification": "40x", "data_structure": { "slide_directories": "Slide_XX (original images)", "gt_directories": "GT_XX (ground truth masks)", "pairing": "same_filename_in_corresponding_directories" }, "processing_notes": [ "Binary ground truth masks converted to instance segmentation", "Connected components analysis used for instance separation", "Small components filtered based on area thresholds", "Images normalized to [0, 1] range" ] }这种数据集能直接用图片替代吗
08-06
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值