Infrared Normalized Difference Vegetation Index for Sentinel-2A Imagery

Abstract

Normalized Difference Vegetation Index (NDVI) is widely applied in several applications. It can be used to classify or identify the land coverage. However, the haze and cloud often disturb the accuracy of this value, which decreases the application performance. Thanks to the electromagnetic wave in the infrared bands can through some atmospheric condition types, and it can employ to estimate the visible bands. Therefore, we propose the Infrared NDVI (INDVI) for Sentinel-2A image applications that tolerate the atmospheric conditions. This index is formulated from the traditional NDVI definition by replacing the red band with the synthetic red band. The linear combination of the infrared bands is employed to construct the synthetic red band formula. The multiple linear regression is used to determine the linear coefficients by using clear image samples. Thus, INDVI formula is calculated via the infrared information only. The degraded Sentinel-2A images by two haze types are used to evaluate our proposed method performance. The INDVI is compared with actual-like NDVI from the clear image that has the nearest sensing date. Results show that our proposed method can give the INDVI value corresponds and close to the standard NDVI, especially in the vegetation coverage area. Moreover, it can be used to approximately determine the NDVI value, although the considered image is full coverage by the haze.