**End**

By applying the LocalizedCorrection algorithm, the *ICLS* will be corrected segmen<sup>t</sup> by segmen<sup>t</sup> through the CRF algorithm.

#### *2.3. Overall Process of the End-to-End and Localized Post-Processing Method*

Based on the four steps and algorithms described in the preceding subsection, we can evaluate the classification result image without requiring testing samples and correct the classification result image within local areas. By integrating these algorithms, we propose an end-to-end and localized post-processing method (ELP) whose input is a remote sensing image *Iimage* and a classification result image *Icls*, and whose output is the corrected classification result image. Through the iterative and

progressive correction process, the goal of improving the quality of the *ICLS* can be achieved. The process of ELP is shown in Figure 3.

**Figure 3.** Overall processes of end-to-end and localized post-processing method (ELP).

As Figure 3 shows, the ELP method is a step-by-step iterative correction process that requires a total of γ iterations to correct the *Icls* content. Before beginning the iteration, the ELP method obtains the segmentation result image *ISLIC* before iteration:

$$I\_{SLIC} = SLIC(I\_{\text{image}}),\tag{7}$$

Then, it evaluates the segments and constructs the suspicion evaluation list for the segments *HList*:

$$\text{HList} = \text{Suspricions} \text{Contruction} (I\_{\text{SLIC}}), \tag{8}$$

In each iteration, ELP updates *HList* to obtain *HList*η, and it outputs a new classification result image *<sup>I</sup>*η*CLS*, where η is the iteration value (in the range [1,γ]). The *i*-th iteration's output is:

$$HList^i = Sensitivity\,\text{Evalulation}(HList^{i-1}, l\_{cls}^{i-1}),\tag{9}$$

$$I\_{\rm cls}^{i} = \text{LocalizedCorrection}(I\_{\rm cls}^{i-1}, \text{HList}^{i}), \tag{10}$$

When η = 1, *HList*<sup>0</sup> = *HList*, and *I*0*cls* = *Icls*; when η ≥ 2, the current iteration result depends on the result of the previous iteration. Based on the above two formulas, the ELP algorithm will update *HList*

and *Icls* in each iteration; *HList* indicates suspicious areas, and these areas are corrected and stored in *Icls*. As the iteration progresses, the final result is obtained:

$$PostProcessResult = I\_{cls}^{\gamma} \tag{11}$$

Through the above process, ELP achieves both the desired goals: end-to-end result image evaluation and localized post-processing.
